Date: Sun, 2 Nov 1997 14:18:15 +0200 (IST)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: John Machin <sjmachin AT lexicon DOT net DOT au>
cc: "Salvador Eduardo Tropea (SET)" <salvador AT inti DOT edu DOT ar>, djgpp AT delorie DOT com
Subject: Re: malloc()
In-Reply-To: <199710302300.KAA01562@mona.lexicon.net.au>
Message-ID: <Pine.SUN.3.91.971102141749.11098F-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk


On Fri, 31 Oct 1997, John Machin wrote:

> (b) The DL's overhead is ranges from about 5% to 20%. The BSD's 
> overhead is over 50%. Note: 50% overhead means for every 100 bytes 
> that I've malloc'ed but not free'd, the package has grabbed 150 
> bytes, i.e 50 bytes wasted.
> (c) The BSD version runs out of real memory & starts swapping sooner.
> (d) The BSD version runs out of swap memory & fails sooner.

It is very important to know which DPMI server was used, and test this
with different DPMI servers.  Each one of them manages memory
differently, and has different bugs/misfeatures that affect when
paging begins.

In addition, each DPMI server might have its own policies as far as
subdivision of memory into chunks is concerned, and that might also
affect the test results.

Furthermore, the DJGPP `sbrk' function (defined on crt0.S in the
library sources) has something to say about memory chunking as well.
It just might happen that it assumes something about `malloc' in the
library.

Anyway, I think this thread should move to djgpp-workers (write to
djgpp-workers AT delorie DOT com), which will ensure that DJ, Charles
Sandmann, Morten Welinder, and a few others who know a lot about these
issues and can contribute, don't need to wade through the news group
to see these messages.